Secondary Logo

Journal Logo

EPIDEMIOLOGY AND SOCIAL: CONCISE COMMUNICATIONS

Revealing HIV epidemic dynamics and contrasting responses in two WHO Eastern European countries: insights from modeling and data triangulation

Marty, Lisea; Lemsalu, Liisb,∗; Ķīvīte-Urtāne, Andac,∗; Costagliola, Dominiquea; Kaupe, Rutac,d; Linina, Indrac; Upmace, Ingac,e; Rüütel, Kristib; Supervie, Virginiea; the HERMETIC study group†

Author Information
doi: 10.1097/QAD.0000000000002778

Abstract

Introduction

The eastern part of the WHO European Region, where HIV rapidly spread among people who inject drugs (PWID) in the late 1990s [1–3], represents one of the few regions globally where HIV epidemics are still growing [4]. From 2009 to 2018, rates of new HIV diagnoses increased by 30% in this region [5] and current coverages of harm reduction services and ART remain too low in many countries to control HIV transmission [6].

In the early 2000s, Estonia and Latvia had, with Russia, the highest rates of new HIV diagnoses in the region [7]. Throughout the 2000s, the number of cases due to sexual transmission increased in several countries in the region [5], including Estonia and Latvia [1,8], raising concerns about a possible shift from PWID- to sexually driven transmission. In the 2010s, although Estonia has reported a decline in new diagnoses rates, this was not the case for Latvia, and many countries in the region [9].

Here, we analyzed HIV situations in Estonia and Latvia, two neighboring countries with marked similarities in initial epidemic characteristics, to understand epidemic dynamics and the impact of HIV response on these dynamics. Specifically, we estimated epidemiological indicators, and then triangulated these estimates with programmatic data to apprehend epidemic trajectories, and identified current gaps in prevention and care.

Methods

Using national surveillance data on newly diagnosed HIV cases in Estonia and Latvia and a clinical stage-based back-calculation model [10–12], we estimated numbers of new and undiagnosed HIV infections and distribution of time from infection to diagnosis; see details in the Supplementary Material (SM).

Briefly, data on new diagnoses, including diagnosis date, sex, exposure group, and clinical stage (derived from International Classification of Diseases-10 code), were available until 2016, at the time of the analysis, and required using multiple imputation techniques to treat missing values (SM, Sections S2-S3, Tables S1 and S2, https://links.lww.com/QAD/B929).

The back-calculation model simultaneously estimates trends in numbers of new HIV infections (i.e., incidence) and in the distribution of times from infection to diagnosis by fitting observed trends in newly diagnosed HIV cases (Figure S1, https://links.lww.com/QAD/B929), stratified by clinical stage at diagnosis (i.e., recent infection, AIDS, neither AIDS nor recent infection). The natural distribution of times from infection to AIDS and the distribution of times from infection to diagnosis for individuals diagnosed with a recent infection were assumed to be known, with respective median values of 10 years and 3 months. The distribution of times from infection to diagnosis for individuals diagnosed without AIDS, nor recent infection, depended on two unknown parameters, representing uptake of routine testing and onset of pre-AIDS related symptoms, which were estimated together with incidence during the estimation procedure. Then, estimated numbers of new infections are projected forward according to estimated distributions of time to diagnosis to obtain numbers of undiagnosed infections in a given year.

Rates of HIV incidence and undiagnosed infections were calculated using population sizes, which were estimated using national statistics and surveys on sexual behavior and drug use (SM, Sections S5-S6, Tables S3-S5, https://links.lww.com/QAD/B929).

To analyze prevention and care programs, we reviewed data from national reports, bio-behavioral surveys, and national public health authorities on the coverage of the following four services: HIV testing, ART, needle and syringe programs (NSP), and opioid agonist treatment (OAT; SM, Sections S8-S9).

All data and estimates were produced at national level, by sex and exposure group, that is, PWID, men who have sex with men (MSM), and heterosexuals, whenever possible and appropriate.

Results

From 2007 to 2016, HIV incidence decreased in Estonia by 61% overall, particularly for male PWID (97%), but also for other exposure groups, except MSM, for which it increased by 418% (Figure 1, Table S6, https://links.lww.com/QAD/B929). In contrast, in Latvia, it increased by 72% overall, particularly for female PWID (121%) and heterosexual women (90%). In 2016, the incidence rate was almost twice higher in Latvia than in Estonia: 35 versus 19 per 100 000 (P < 0.05). In both countries, PWID were the most affected population in terms of incidence rate, however most new infections occurred via heterosexual contacts; 50% in Latvia and 75% in Estonia.

F1
Fig. 1:
Estimated annual number of new HIV infections in Estonia (circles) and in Latvia (triangles) over 2007-2016, overall (a) and by exposure group, female persons who inject drugs (PWID) (b), male PWID (c), men who have sex with men, (d), heterosexual women (e), and heterosexual men (f).

In both countries, the estimated median time from infection to diagnosis was shorter for PWID (3.1 years in Estonia, 3.4 years in Latvia, Figure S2, https://links.lww.com/QAD/B929) than for other exposure groups (all P values < 0.05). In Estonia, MSM had a shorter median time to diagnosis than heterosexual women and men (respectively 3.4, 4.1, and 4.7, all P values < 0.05), whereas in Latvia, MSM had a longer median time than heterosexual women and shorter than heterosexual men (respectively 4.4, 3.8, and 4.9, all P values < 0.05). Over time, median times to diagnosis considerably decreased for male PWID in Estonia, from 3.5 to 2.6 years (P < 0.05, Figure S3, https://links.lww.com/QAD/B929).

In 2016, the rate of undiagnosed infections was higher in Latvia than in Estonia (135 versus 107 per 100 000, P < 0.05, Table S7, https://links.lww.com/QAD/B929). PWID were the most affected population in terms of undiagnosed infection rate, however the majority of individuals living with undiagnosed HIV were heterosexual; 52% in Latvia and 76% in Estonia.

Coverage of harm reduction, HIV testing, and ART was higher in Estonia than in Latvia (Figure 2). In Estonia, the number of needles and syringes distributed annually and the number of OAT clients increased rapidly in the late 2000s, and then remained rather stable, whereas in Latvia, they increased more recently and reached lower levels than in Estonia. Relative to the estimated number of PWID, in Estonia, coverage was 224 needles and syringes distributed per PWID and 13 OAT clients per 100 PWID in 2016, and respectively 94 and 8 in Latvia.

F2
Fig. 2:
Harm reduction services (a, b), HIV testing coverage (c–e), and antiretroviral treatment (ART) coverage (f, g) in Estonia (circles) and Latvia (triangles) from 2000 to 2016.

In Estonia, the proportion of individuals tested for HIV within the past 12 months were much higher among PWID and MSM (respectively, 50% and 43% in 2016) than in the general population (13% in 2014, last data point available). In Latvia, these proportions remained low over time.

ART coverage among individuals diagnosed with HIV increased in both countries, but reached a higher level in Estonia, particularly among PWID (all P values < 0.05). No ART data are shown for MSM as existing studies were limited by a small sample size (SM, sections S8-S9).

Discussion

Our findings show that Estonia has turned the tide of HIV, particularly for male PWID, for which both incidence and time to diagnosis considerably decreased over time, whereas in Latvia the epidemic remains very active. These epidemic trajectories mirror differences in harm reduction, particularly NSP, HIV testing, and ART coverages between the two countries, especially for PWID. In Estonia, these coverages have expanded over time and reached medium to high levels, whereas in Latvia, they remained at rather low levels. This has probably contributed to reducing HIV transmission among PWID in Estonia, whereas in Latvia, it remained too low to control it.

The decrease in HIV incidence for PWID in Estonia has been concomitant with a reduction of HIV acquisition via heterosexual contacts, whereas in Latvia, acquisition via heterosexual contacts has tended to increase recently. HIV acquisition via heterosexual contacts may have been influenced by the epidemic trajectory and ART coverage among PWID. Indeed, PWID are sexually active, many do not always use condoms [13], and some engage in sex work [14,15], thus sexual transmission from HIV-infected PWID to their sexual partners can occur; several studies showed that, in Latvia, the heterosexual epidemic had multiple interactions with the PWID epidemic [8,16,17]. Hence, in Estonia, sexual partners of PWID could have been less likely to acquire HIV over time, as ART coverage increased and HIV incidence decreased among PWID, whereas in Latvia, sexual transmission possibly remained high due to low ART coverage and increasing incidence among PWID over time.

Currently, in both countries, and despite the remarkable reduction in incidence in Estonia, PWID remains the most affected exposure group in terms of rates of HIV incidence and undiagnosed infections. However, it is worth noticing that half of the new and undiagnosed infections in Latvia and three-quarters in Estonia are now affecting individuals who report heterosexual exposure. This constitutes a shift compared to the initial epidemiological situation in both countries, in which the majority of cases were attributed to injecting-drug use [7]. We hypothesize that, in Latvia, this shift occurred because of the failure to control HIV among PWID and the subsequent spread of HIV from PWID to their sexual partners, whereas in Estonia, it occurred because transmission via injecting-drug use was further controlled than transmission via heterosexual contacts, which was also controlled but to a lesser extent.

Of important note, we found that HIV incidence among MSM increased in both countries, and thus, prevention, testing, and ART coverage are probably insufficient to control HIV spread in this group. We found that existing data for MSM on testing in Latvia and on ART in both countries were rather scarce and limited [18].

Some of our results are consistent with previous findings (SM, section S11), however, our study has limitations (see [19] for limitations regarding the back-calculation model). Recent trends inferred for heterosexual women in Latvia need to be confirmed as confidence intervals were large. The estimates’ accuracy may have been limited by several sources of uncertainty. Incidence may have been underestimated due to pre-HIV diagnosis mortality, especially for PWID where mortality is high. Potential inaccurate adjustment for missing entries (e.g., ∼50% missing for exposure group in Estonia) might have affected our estimates. Stigma and discrimination against MSM and PWID may have led some individuals to avoid reporting any HIV exposure risk or to report heterosexual exposure, instead of reporting exposure via male-to-male sexual contact or injecting-drug use. Hence, we may have overestimated transmission via heterosexual contact, and in turn, the shift toward heterosexual route as main transmission mode, and underestimated transmission via injecting-drug use and male-to-male sexual contact. In addition, we relied on sub-national data for testing and ART for PWID in Estonia, which may not be representative of the national coverage. Finally, ecological correlation does not imply causation, and disentangling the impact of interventions from other factors remains challenging, therefore, our findings linking HIV incidence decrease with increase in harm reduction and ART coverage for PWID in Estonia should be taken with caution. Indeed, besides the scale-up of HIV interventions, other factors may have influenced and interfered with HIV spread, including natural waning of transmission induced by a saturation effect [14,20,21], local heterogeneities [8], and social and spatial structure of PWID populations [22,23].

Conclusions

Although HIV epidemics are worsening in many countries of the Eastern part of the WHO European Region, including Russia [24], our findings show that, in Estonia, HIV transmission was considerably reduced among PWID, and to a lesser extent among heterosexuals, whereas in Latvia, it remains uncontrolled. A decrease in incidence in Estonia was concomitant with an increase in harm reduction and ART coverage for PWID. Several challenges remain, however, to be faced to control or further control epidemics in both countries. Tailored prevention (e.g., preexposure prophylaxis) and testing services for MSM are limited and need to be expanded, as well as OAT for PWID and prevention interventions for their sexual partners. Furthermore, in Latvia, scale-up of harm reduction and ART for PWID is urgently needed to avoid further expansion of the epidemic.

Acknowledgements

LM and VS conceived the model. LM conducted the numerical simulations. All authors participated in the writing of the manuscript, analysis, and interpretation of the results. All authors have read and approved the final manuscript.

This work was supported by the ANRS (France) and the State Education Development Agency, Republic of Latvia (Latvia) through the framework of HIVERA JTC 2014; and in Estonia by the Estonian Ministry of Education and Research (IUT 42-2) and the National Health Plan 2009-2020. LM is grateful for the financial support of the ANRS, in the form of a postdoctoral research fellowship.

Authors would like to express special gratitude to the Centre for Disease Prevention and Control of Latvia (and personally to Sarlote Konova and Kristine Ozolina), to the National Health Service of Latvia, to Liilia Lõhmus, Sigrid Vorobjov, Maris Salekešin, and Kaire Vals from the National Institute for Health Development, Estonia, for their responsiveness and provision of data for mathematical modeling and triangulation exercises of the current study, and to the HERMETIC Study Group.

Conflicts of interest

There are no conflicts of interest.

References

1. Donoghoe MC, Lazarus JV, Matic S. HIV/AIDS in the transitional countries of eastern Europe and central Asia. Clin Med 2005; 5:487–490.
2. Uusküla DA, Rajaleid K, Talu A, Abel K, Rüütel K, Hay G. Estimating injection drug use prevalence using state wide administrative data sources: Estonia, 2004. Addict ResTheory 2007; 15:411–424.
3. Dehne KL, Khodakevich L, Hamers FF, Schwartländer B. The HIV/AIDS epidemic in eastern Europe: recent patterns and trends and their implications for policy-making. AIDS 1999; 13:741–749.
4. Case KK, Johnson LF, Mahy M, Marsh K, Supervie V, Eaton JW. Summarizing the results and methods of the 2019 Joint United Nations Programme on HIV/AIDS HIV estimates. AIDS 2019; 33: (suppl 3): S197–S201.
5. European Center for Disease Prevention and Control/WHO Regional Office. HIV/AIDS surveillance in Europe 2019–2018 data
6. Saldanha VP, Buse K. AIDS in eastern Europe and central Asia: time to face the facts. Lancet 2016; 388:2737–2738.
7. Hamers FF, Downs AM. HIV in central and eastern Europe. Lancet 2003; 361:1035–1044.
8. Graw F, Leitner T, Ribeiro RM. Agent-based and phylogenetic analyses reveal how HIV-1 moves between risk groups: injecting drug users sustain the heterosexual epidemic in Latvia. Epidemics 2012; 4:104–116.
9. European Center for Disease Prevention and Control/WHO Regional Office. HIV/AIDS surveillance in Europe 2017–2016 data
10. Ndawinz JDA, Costagliola D, Supervie V. New method for estimating HIV incidence and time from infection to diagnosis using HIV surveillance data: results for France. AIDS 2011; 25:1905–1913.
11. Supervie V, Ndawinz JDA, Lodi S, Costagliola D. The undiagnosed HIV epidemic in France and its implications for HIV screening strategies. AIDS 2014; 28:1797–1804.
12. Marty L, Cazein F, Panjo H, Pillonel J, Costagliola D, Supervie V. Revealing geographical and population heterogeneity in HIV incidence, undiagnosed HIV prevalence and time to diagnosis to improve prevention and care: estimates for France. J Int AIDS Soc 2018; 21:e25100.
13. Rüütel K, Trummal A, Salekešin M, Pervilhac C. HIV Epidemic in Estonia: Analysis of Strategic Information. Case Study. Copenhagen, Denmark: World Health Organization Regional Office for Europe; 2011.
14. Vorobjov S, Salekešin K. HIVi levimus ja riskikäitumine narkootikume süstivate inimeste hulgas, Kohtla-Järve 2016 [HIV prevalence and risk behaviours among people who inject drugs in Kohtla-Järve 2016]. Tallinn: National Institute for Health Development; 2017. https://www.tai.ee/et/terviseandmed/uuringud/download/401. [accessed 30 November 2020]
15. Ķīvīte-Urtāne A, Civjāne L, Isajeva L, Kaupe R, Vanaga-Araja D, Veisberga L. Narkotiku lietošanas paradumi un tendences Latvijā. Narkotiku lietotāju kohortas pētījuma 11. posma rezultāti. Rīga: Slimību profilakses un kontroles centrs, DIA+LOGS; 2019. https://www.spkc.gov.lv/lv/petijumi/narkotiku-lietosanas-paradumi-un-tendences-latvija-2019.pdf [accessed 30 November 2020]
16. Balode D, Skar H, Mild M, Kolupajeva T, Ferdats A, Rozentale B, et al. Phylogenetic analysis of the Latvian HIV-1 epidemic. AIDS Res Hum Retroviruses 2012; 28:928–932.
17. Stadler T, Bonhoeffer S. Uncovering epidemiological dynamics in heterogeneous host populations using phylogenetic methods. Philos Trans R Soc Lond B Biol Sci 2013; 368:20120198.
18. Lutsar I, Avi R, Soodla P, Huik K, Toompere K, Jääger M, et al. UNAIDSi kriteeriumid HIV-infektsiooni monitoorimiseks: olukord Eestis 2017. aastal E-HIVi põhjal. Eesti Arst 2018; 97:527–534.
19. Marty L, Van Beckhoven D, Ost C, Deblonde J, Costagliola D, Sasse A, et al. Estimates of the HIV undiagnosed population in Belgium reveals higher prevalence for MSM with foreign nationality and for geographic areas hosting big cities. J Int AIDS Soc 2019; 22:e25371.
20. Vickerman P, Hickman M. Commentary on de Vos et al. (2013): can ecological trends in HIV or HCV incidence be used to assess intervention impact?. Addiction 2013; 108:1082–1083.
21. Vorobjov S, Rüütel K, Abel-Ollo K, Salekešin K. HIVi levimuse ja riskikäitumise uuring Narva süstivate narkomaanide seas 2014 [HIV prevalence and risk behaviours among people who inject drugs in Narva 2014]. Tallinn: National Institute for Health Development; 2015. https://www.tai.ee/et/terviseandmed/uuringud/download/318. [accessed 30 November 2020]
22. Uusküla A, Rajaleid K, Talu A, Abel-Ollo K, Jarlais DCD. A decline in the prevalence of injecting drug users in Estonia, 2005-2009. Int J Drug Policy 2013; 24:312–318.
23. Raag M, Vorobjov S, Uusküla A. Prevalence of injecting drug use in Estonia 2010-2015: a capture-recapture study. Harm Reduct J 2019; 16:19.
24. Beyrer C, Wirtz AL, O’Hara G, Léon N, Kazatchkine M. The expanding epidemic of HIV-1 in the Russian Federation. PLoS Med 2017; 14:e1002462.

Liis Lemsalu and Anda Kivite-Urtane: Both authors equally contributed to this work.

HERMETIC Study Group: Hanne Apers (ITM, Belgium), Jessika Deblonde (Sciensano, Belgium), Anda Ķīvīte-Urtāne (RSU, Latvia), Jasna Loos (ITM, Belgium), Lise Marty (INSERM U1136, France), Christiana Nöstlinger (ITM, Belgium), Daniela Rojas Castro (Coalition Plus, France), Virginie Supervie (INSERM U1136, France), Dominique Van Beckhoven (Sciensano, Belgium).

Keywords:

Eastern Europe; harm reduction; HIV; men who have sex with men; PWID

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.